Gallup has now started reporting their results in three formats: all registered voters (RV); likely voters according to traditional methodology that filters out people who have not voted in previous presidential cycles to large degrees (LV-traditional); and a likely voter screen that is based on reported intent to vote without consideration to past voting behaviors (LV-expanded).
The results are as expected: traditional LV screens take off 4 points from the results of all RVs (in today’s case 10 becomes 6) but a LV screen based primarily on intent to vote, one that captures the newly registered and newly empowered voters who are overwhelmingly for Obama, gives results 4 or 5 points higher than the traditional LV screen.
Why does this matter?
Many state polls use traditional LV screens. For a state with an average distribution of newly registered/empowered voters these polls may be underestimating Obama’s numbers by 4 or 5 points by so doing (if these voters do indeed come out this time). In states that have larger numbers of newly registered/empowered voters … say for example a large Hispanic population that has been quite energized if their primary behavior is representative … that underestimation may be significantly more sizable.
Bradley effect Schmadley effect. Between the underestimation inherent in the traditional LV screen, the cell phone effect, the ground game, and some depression among some of the GOP core … some very traditional Red states may be flipping Blue when all is said and done.